3 ms·

Computers vs. Brains

10 points by robg 18y ago

warfangle 18y ago
"To put this in perspective, the entire archived contents of the Internet fill just three petabytes." This is categorically false. While the "Internet Archive" currently has the capacity of 4.5 petabytes, the sheer amount of data available on the http protocol is more likely in exabytes (especially once you take into consideration high-capacity services like youtube, bittorrent, and other file sharing protocols).
niels_olson 18y ago
um, I rather doubt a computer is going to consume a gigawatt in 15 years.
swombat 18y ago
Wow, way to spectacularly miss the point! 1) No intelligent person is suggesting that a computer will directly map to a human brain. We're not trying to reproduce a brain in silicon. However, a sufficiently powerful computer might be able to model the workings of a brain - or even might not need to if we find another path to intelligence. 2) If anything, the complexity and existence of the human brain is an indication that such a system can be built efficiently. After all, the brain exists, therefore it should be possible to at least build something of comparable complexity and power, at some point in the future. Maybe it won't be as rapid as Moore's law predicts, but there's no reason to think that we won't eventually be able to reproduce what nature has been able to do through blind selection. 3) There's no reason why a singularity can only come via wholly artificial intelligence. Intelligence augmentation could easily achieve the same objective (and is perhaps even more likely - after all, it's happening today already, albeit on a smaller scale). Did they even bother to read up on Singularity stuff before writing this article?
- rjprins 18y ago
  Agreed! The author obviously has not thought a lot about AI before and misses the idea of abstract models/functions as implemented by brains. And she is wrong on the "dark, hot underbelly of Moore's law", power consumption relative to computational power has most certainly gone down in an exponential fashion.
- robg 18y ago
  Your criticisms are fair, but: 1) http://en.wikipedia.org/wiki/Blue_Brain http://en.wikipedia.org/wiki/Blue_Brain 2) How long do you think we'll need? "At some point in the future" is an awfully long time scale, especially relative to how long it took natural selection to create the human brain. 3) I don't think they'd deny that augmentation is already happening. But that piecemeal progress is still a long way from The Singularity.
  - swombat 18y ago
    1) is exactly what I suggested - a virtual model of the brain. I don't see any "but" there... 2) I'm no AI expert, so my opinion of this is pretty irrelevant, but I would be extremely surprised if we didn't have some form of human-level intelligence within the next 100 years, and I would not be surprised if we had it in the next 20 years. 3) Piecemeal progress that feeds back on itself is the essence of the whole Singularity hypothesis. Better brains come up with even better ways to enhance themselves, which makes them even better, etc. Technological progress can grow exponentially in ways not subject to the Malthusian limits.
varjag 18y ago
Give us the million years evolution had to develop the brain, we would surely come up with something better.
danbmil99 18y ago
Why do neuroscientists get so hot & bothered over the prospect of AI? It doesn't diminish the brain's complexity to think that one day we can rival it with alternate technology. There's a whiff of duality/magic animus in this line of thinking.
dejb 18y ago
Be warned you may lose several IQ points reading this article if you are familiar with the concept of strong AI and the singularity. Short version - - Strong AI isn't possible because it will require too much power according to Moore's law. - Even if we could build Strong AI, why bother?
ilaksh 18y ago
This points out how strong status quo bias is even for neuroscientists. Which is an indication that we may be in for a very significant collision between reality (artificial general intelligence) and mainstream humanity. I think that it is going to be quite some time (years or maybe decades) before most people accept and comprehend the changes brought on by AGI and start to modify social and other systems to adjust to this reality.

10 points by robg 18y ago

"To put this in perspective, the entire archived contents of the Internet fill just three petabytes." This is categorically false. While the "Internet Archive" currently has the capacity of 4.5 petabytes, the sheer amount of data available on the http protocol is more likely in exabytes (especially once you take into consideration high-capacity services like youtube, bittorrent, and other file sharing protocols).

niels_olson 18y ago

um, I rather doubt a computer is going to consume a gigawatt in 15 years.

swombat 18y ago

Wow, way to spectacularly miss the point! 1) No intelligent person is suggesting that a computer will directly map to a human brain. We're not trying to reproduce a brain in silicon. However, a sufficiently powerful computer might be able to model the workings of a brain - or even might not need to if we find another path to intelligence. 2) If anything, the complexity and existence of the human brain is an indication that such a system can be built efficiently. After all, the brain exists, therefore it should be possible to at least build something of comparable complexity and power, at some point in the future. Maybe it won't be as rapid as Moore's law predicts, but there's no reason to think that we won't eventually be able to reproduce what nature has been able to do through blind selection. 3) There's no reason why a singularity can only come via wholly artificial intelligence. Intelligence augmentation could easily achieve the same objective (and is perhaps even more likely - after all, it's happening today already, albeit on a smaller scale). Did they even bother to read up on Singularity stuff before writing this article?

rjprins 18y ago
Agreed! The author obviously has not thought a lot about AI before and misses the idea of abstract models/functions as implemented by brains. And she is wrong on the "dark, hot underbelly of Moore's law", power consumption relative to computational power has most certainly gone down in an exponential fashion.
robg 18y ago
Your criticisms are fair, but: 1) http://en.wikipedia.org/wiki/Blue_Brain http://en.wikipedia.org/wiki/Blue_Brain 2) How long do you think we'll need? "At some point in the future" is an awfully long time scale, especially relative to how long it took natural selection to create the human brain. 3) I don't think they'd deny that augmentation is already happening. But that piecemeal progress is still a long way from The Singularity.
- swombat 18y ago
  1) is exactly what I suggested - a virtual model of the brain. I don't see any "but" there... 2) I'm no AI expert, so my opinion of this is pretty irrelevant, but I would be extremely surprised if we didn't have some form of human-level intelligence within the next 100 years, and I would not be surprised if we had it in the next 20 years. 3) Piecemeal progress that feeds back on itself is the essence of the whole Singularity hypothesis. Better brains come up with even better ways to enhance themselves, which makes them even better, etc. Technological progress can grow exponentially in ways not subject to the Malthusian limits.

varjag 18y ago

Give us the million years evolution had to develop the brain, we would surely come up with something better.

danbmil99 18y ago

Why do neuroscientists get so hot & bothered over the prospect of AI? It doesn't diminish the brain's complexity to think that one day we can rival it with alternate technology. There's a whiff of duality/magic animus in this line of thinking.

dejb 18y ago

Be warned you may lose several IQ points reading this article if you are familiar with the concept of strong AI and the singularity. Short version - - Strong AI isn't possible because it will require too much power according to Moore's law. - Even if we could build Strong AI, why bother?

ilaksh 18y ago

This points out how strong status quo bias is even for neuroscientists. Which is an indication that we may be in for a very significant collision between reality (artificial general intelligence) and mainstream humanity. I think that it is going to be quite some time (years or maybe decades) before most people accept and comprehend the changes brought on by AGI and start to modify social and other systems to adjust to this reality.